strange clusters from stranded mRNA library

[JBKri]

We recently had a failed MiSeq run from an mRNA library (TruSeq stranded mRNA). Initially we though it was just a matter of overclustering, but on closer inspection we noticed the clusters looked strange. We did not get any clear, starlike clusters, instead we got small "specks" and much lower FWHM than normal. Upon rechecking the qPCR results we saw that we had two peaks in the melting curve. Thinking we might have primer-adapter dimers, we did two extra Ampure cleanups of the libraries and ran the products on our Bioanalyser and did another qPCR. The size distribution is now narrower than before, and the secondary peak in the melting curves is smaller, but still present. Is it normal to have two peaks in the melting curve? The libraries are all from a single species and tissue. We don't understand what went wrong with these libraries, and now we are uncertain if we should try another run with the extra cleanup steps, or repeat the library preps from the beginning.
Jon

[nucacidhunter]

I do not see any issues with your original library except following:
1- low yield if you have run library without dilution on BA
2- large insert sizes which is not typical for TruSeq RNA libraries

By doing extera clean up you have lost part of smaller fragments.

Melt curve analysis is not essential for library quantification as library is composed of heterogeneous fragments and more than one peak in melt curve is expected.

You have not given sufficient information on run faileur but over clustering is a possibility.

[JBKri]

I do not see any issues with your original library except following:
1- low yield if you have run library without dilution on BA
2- large insert sizes which is not typical for TruSeq RNA libraries

By doing extera clean up you have lost part of smaller fragments.

Melt curve analysis is not essential for library quantification as library is composed of heterogeneous fragments and more than one peak in melt curve is expected.

You have not given sufficient information on run faileur but over clustering is a possibility.

This was a v3-600 run. The cluster density was reported as 1253 k/mm2, but Illumina tech support suggested this was an underestimate due to overclustering. 0 % passed filtering. I attached a few more images showing various statistics.

1. The yield was good, these were diluted.
2. We did zero fragmentation time, so we expected the large inserts. We have done two previous successful libraries and runs with the exact same protocol.

[nucacidhunter]

I tend to agree with Illumina tech support re overclustering. If you are happy with this type of library then you need to repeat run with less input for clustering. Overclustering could have been result of underestimating library concentration or somehow more than desired amount of library has been used for clustering such as human error or faulty pipette.